Russian scientists have developed a graphical model that allows for the analysis of neurophysiological and genetic data in order to identify depression.
This was stated in the bulletin "Science in Siberia" issued by the Siberian branch of the Russian Academy of Sciences.
The bulletin quoted one of the developers, Alexander Savostyanov, as saying, "The new method of data processing is based on a graph system, which is a type of neural network. The method works as follows: First, we use what is called a training sample, where we show the healthy state versus the depressive state, and we load their genetic patterns and electroencephalogram (EEG) data, and the neural network learns to distinguish between the two states. Then we create a test sample containing only the genetic patterns and EEG data, and the neural network must guess who is sick and who is healthy."
Scientists were able to collect a large dataset covering almost all of Siberia, from the Altai region to Buryatia. They examined more than 3,000 people. Blood samples and oral epithelial swabs (cheek mucosa scrapings) were taken from some of the participants, and they underwent electroencephalograms (EEGs) and psychological questionnaires to determine their psychological characteristics, including the severity of depressive symptoms.
For DNA extracted from blood and cheek swabs, 164 gene loci were sequenced, and based on this, each person's genetic deviations from the so-called reference genome were identified.
The accuracy of the graphical approach proved significantly higher than that of previously tested neural network structures. While the accuracy of depression detection was previously around 86%, it reached over 96% here. Furthermore, the neural network produced fewer false negatives compared to previously tested models, which is particularly important in the context of initial screening for mental illness, where the cost of such errors can be high. Failing to detect depression in someone who actually has it is far more serious than misdiagnosing a non-existent case of depression.
Savostyanov noted, "A person's genetic predisposition to depression can be determined from birth, but the problem is that if only the genetic profile is used, the accuracy of the predictions will be very low, as life circumstances have a significant impact on the development of depression. Electroencephalography (EEG) is good at assessing the condition of a person being examined at a specific moment, but this indicator changes very rapidly and does not reflect the person's long-term characteristics, and may be affected by hunger or other physical conditions. In our practice, there was a case where the EEG analysis showed an abnormality in the examinee's brain, and it turned out that this condition resulted from the person coming in for the EEG immediately after having a tooth extracted. When genetic, neurophysiological, and behavioral data are analyzed, the reliability of the diagnosis increases considerably."
